Monitoring NiFi – Workflow SLA

Note – This article is part of a series discussing subjects around NiFi monitoring.

Depending of the workflows and use cases, you may want to retrieve some metrics per use-case/workflow instead of high level metrics. One of the things you could be looking after is Workflow SLA.

First of all, we need to agree on what is “Workflow SLA”. In this article, by SLA (Service-Level Agreement) monitoring I mean that I want to ensure that a workflow is processing every single event quickly enough. In other words, if an event took a long time to be processed, I want to be aware of that so I can understand what is going on. A workflow could be perfectly running in NiFi (no bulletin generated, no error) but the processing time could get bigger because of external causes (resources exhaustion, network bandwidth with external systems, abnormal event size, etc).

Luckily for us, every single flow file within NiFi contains two core attributes that are perfectly designed for our needs. From the documentation:

  • entryDate: The date and time at which the FlowFile entered the system (i.e., was created). The value of this attribute is a number that represents the number of milliseconds since midnight, Jan. 1, 1970 (UTC).
  • lineageStartDate: Any time that a FlowFile is cloned, merged, or split, this results in a “child” FlowFile being created. As those children are then cloned, merged, or split, a chain of ancestors is built. This value represents the date and time at which the oldest ancestor entered the system. Another way to think about this is that this attribute represents the latency of the FlowFile through the system. The value is a number that represents the number of milliseconds since midnight, Jan. 1, 1970 (UTC).

In our case we are really interested by the “lineageStartDate” attribute but based on your specific use cases and needs you could also be interested by “entryDate”.

At the end of my workflow I want to monitor, I could use a RouteOnAttribute processor to route the event if and only if the duration since the lineage start date is over a given threshold (let’s say one minute for the below example). And if an event took longer than expected, I’m routing the flow file to a PutEmail processor to send an email and receive a notification.

Screen Shot 2017-05-10 at 3.33.34 PM.png

My RouteOnAttribute configuration:

Screen Shot 2017-05-10 at 3.34.02 PM.png

This way, at a workflow level, you are able to check if the processing time of an event is meeting your requirements. You could obviously follow the same approach to check the processing time at multiple critical points of the workflow.

Note that, instead of sending an email notification, you could perfectly use any other kind of processor to integrate this alert/event with any system you are using on your side.

One last comment, I recommend you to check out the article about the Ambari Reporting Task and Grafana where I’m also discussing Workflow SLA and how to send the metrics to Ambari Metrics System and display the information into a Grafana dashboard.

As usual feel free to ask questions and comment this post.

Monitoring NiFi – Logback configuration

Note – This article is part of a series discussing subjects around NiFi monitoring.

There is one configuration file in NiFi that manages all the logging operations performed by NiFi: this file is in the configuration directory (NIFI_HOME/conf) and is named logback.xml. It is good to know that this file can be modified “on-the-fly” and it won’t be required to restart NiFi for the modifications to be taken into account.

This file is a common configuration file for the logback library which is a successor of the famous log4j project. Here is the official website. I won’t get into the details of logback itself (documentation is here) but here is a quick overview of the default configuration when using NiFi.

The most important log file in NiFi is certainly nifi-app.log which is created according to this configuration block:

    <appender name="APP_FILE" class="ch.qos.logback.core.rolling.RollingFileAppender">
        <rollingPolicy class="ch.qos.logback.core.rolling.SizeAndTimeBasedRollingPolicy">
        <encoder class="ch.qos.logback.classic.encoder.PatternLayoutEncoder">
            <pattern>%date %level [%thread] %logger{40} %msg%n</pattern>

    <root level="INFO">
        <appender-ref ref="APP_FILE"/>

In this case, log file is rolling every hour or when the file is going over 100MB. Besides we only keep up to 30 files worth of history. Based on your retention policies you can update this file to meet your requirements.

Also note that, by default, you will have the following log files:

  • ./logs/nifi-app.log
  • ./logs/nifi-bootstrap.log
  • ./logs/nifi-user.log

Now… what can we do to use this file for monitoring purpose? Some options here:

  • We can use the TailFile processor of NiFi to let NiFi tails its own log file and perform the required operation when some conditions are met. Example to send an HTTP request when logs of WARN and ERROR levels are detected:

Screen Shot 2017-05-02 at 9.05.14 PM.png

  • We can also define new appenders in the log configuration file and change it according to our needs. In particular, we could be interested by the SMTP Appender that can send logs via emails based on quite a large set of conditions. Full documentation here.
  • Obviously you can also configure this configuration file so that NiFi log files integrate with your existing systems. An idea could be to configure a Syslog appender to also redirect the logs to an external system.

Few remarks regarding the logs as I see recurrent questions:

  • At the moment, there is no option to have a log file per processor or per process group. It is discussed by NIFI-3065. However such an approach could have performance implications since each log message would need to be routed to the correct log file.
  • Also, at the moment, the custom name of the processor is not displayed in the log messages (only the type is used). It is discussed by NIFI-3877. Right now it looks like:

2017-05-12 10:43:37,780 INFO [Timer-Driven Process Thread-4] o.a.nifi.processors.standard.InvokeHTTP InvokeHTTP[id=f7ebb153-015b-1000-f590-599786e16340] my log message…

This could be a solution in a multi-tenant environment (to get one log file per workflow/use case) assuming each “team” is following the same convention to name the components in the workflows.

As usual feel free to ask questions and comment this post.

Monitoring NiFi – Bootstrap notifier

Note – This article is part of a series discussing subjects around NiFi monitoring.

When NiFi is running, there is more than one process running on the server. Let’s have a look:

[root@pvillard-hdf-2 ~]# ps -ef | grep nifi
nifi     19380     1  0 12:33 ?        00:00:00 /bin/sh /usr/hdf/current/nifi/bin/ start
nifi     19382 19380  0 12:33 ?        00:00:06 /usr/java/jdk1.8.0_131//bin/java -cp /usr/hdf/current/nifi/conf:/usr/hdf/current/nifi/lib/bootstrap/* -Xms12m -Xmx24m -Dorg.apache.nifi.bootstrap.config.log.dir=/var/log/nifi -Dorg.apache.nifi.bootstrap.config.file=/usr/hdf/current/nifi/conf/bootstrap.conf org.apache.nifi.bootstrap.RunNiFi start
nifi     19419 19382  7 12:33 ?        00:07:36 /usr/java/jdk1.8.0_131/bin/java -classpath /usr/hdf/current/nifi/conf:/usr/hdf/current/nifi/lib/jcl-over-slf4j-1.7.12.jar:/usr/hdf/current/nifi/lib/jul-to-slf4j-1.7.12.jar:/usr/hdf/current/nifi/lib/log4j-over-slf4j-1.7.12.jar:/usr/hdf/current/nifi/lib/logback-classic-1.1.3.jar:/usr/hdf/current/nifi/lib/logback-core-1.1.3.jar:/usr/hdf/current/nifi/lib/nifi-runtime- -Dorg.apache.jasper.compiler.disablejsr199=true -Xmx512m -Xms512m -Dambari.metrics.collector.url=http://pvillard-hdf-1:6188/ws/v1/timeline/metrics -Djava.awt.headless=true -XX:+UseG1GC -Dnifi.bootstrap.listen.port=39585 -Dapp=NiFi -Dorg.apache.nifi.bootstrap.config.log.dir=/var/log/nifi org.apache.nifi.NiFi -k A2EA52795B33AB2F21C93E7E820D08369F1448478C877F4C710D6E85FD904AE6

As you can see, the script is running a bootstrap (running a JVM between 12 and 24MB) that is in charge of the NiFi JVM itself. In this example, the script is running with the PID 19380 and is the parent of the bootstrap process running with PID 19382. The bootstrap itself is the parent of NiFi running with the PID 19419.

If you only kill the NiFi process, the bootstrap will detect it and automatically relaunch NiFi for you:

kill -9 19419

Then I have:

nifi     19380     1  0 12:33 ?        00:00:00 /bin/sh /usr/hdf/current/nifi/bin/ start
nifi     19382 19380  0 12:33 ?        00:00:06 /usr/java/jdk1.8.0_131//bin/java -cp /usr/hdf/current/nifi/conf:/usr/hdf/current/nifi/lib/bootstrap/* -Xms12m -Xmx24m -Dorg.apache.nifi.bootstrap.config.log.dir=/var/log/nifi -Dorg.apache.nifi.bootstrap.config.file=/usr/hdf/current/nifi/conf/bootstrap.conf org.apache.nifi.bootstrap.RunNiFi start
nifi     24702 19382 99 14:20 ?        00:00:05 /usr/java/jdk1.8.0_131/bin/java -classpath /usr/hdf/current/nifi/conf:/usr/hdf/current/nifi/lib/jcl-over-slf4j-1.7.12.jar:/usr/hdf/current/nifi/lib/jul-to-slf4j-1.7.12.jar:/usr/hdf/current/nifi/lib/log4j-over-slf4j-1.7.12.jar:/usr/hdf/current/nifi/lib/logback-classic-1.1.3.jar:/usr/hdf/current/nifi/lib/logback-core-1.1.3.jar:/usr/hdf/current/nifi/lib/nifi-runtime- -Dorg.apache.jasper.compiler.disablejsr199=true -Xmx512m -Xms512m -Dambari.metrics.collector.url=http://pvillard-hdf-1:6188/ws/v1/timeline/metrics -Djava.awt.headless=true -XX:+UseG1GC -Dnifi.bootstrap.listen.port=39585 -Dapp=NiFi -Dorg.apache.nifi.bootstrap.config.log.dir=/var/log/nifi org.apache.nifi.NiFi -k A2EA52795B33AB2F21C93E7E820D08369F1448478C877F4C710D6E85FD904AE6

As you can see the NiFi process has a new PID since it is a new process launched by the bootstrap.

You can find more information about the bootstrap system in the administration guide. One interesting feature with this bootstrap approach is the bootstrap notifier. It allows you to configure a notification service that will be triggered when the bootstrap starts, stops or detects an interruption of the NiFi process.

With the new version of Apache NiFi (1.2.0), you can send notification to an HTTP(S) endpoint. If you need a custom notification service (example: send SNMP traps), it’s not that hard: you just need to extend AbstractNotificationService. You can have a look at the two existing implementations: email notifier and HTTP notifier.

Let’s configure our NiFi to use the email notification service and see what is the result. On each node of our cluster, we need to update the bootstrap.conf configuration file as below:

# Notification Services for notifying interested parties when NiFi is stopped, started, dies

# XML File that contains the definitions of the notification services

# In the case that we are unable to send a notification for an event, how many times should we retry?

# Comma-separated list of identifiers that are present in the; which services should be used to notify when NiFi is started?

# Comma-separated list of identifiers that are present in the; which services should be used to notify when NiFi is stopped?

# Comma-separated list of identifiers that are present in the; which services should be used to notify when NiFi dies?

In this case, we are saying that the configuration of our notification services (we can define and use multiple notifiers) is in the file


and that we want to use the notifier called “email-notification” (that’s the default name defined in the XML configuration file) for stop, start, and dead events. As stated in the documentation, it’s possible to define a list of notifiers for each type of event.

For this demonstration, I’ll use my Gmail account and the Gmail SMTP server to send the notifications (obviously, with this example, NiFi needs a public internet access to send the requests to the SMTP server). Here is the configuration file:

        <property name="SMTP Hostname"></property>
        <property name="SMTP Port">587</property>
        <property name="SMTP Username"></property>
        <property name="SMTP Password">myPassword</property>
        <property name="SMTP TLS">true</property>
        <property name="From">"NiFi Service Notifier"</property>
        <property name="To">email address that will receive the notifications</property>

Here are the emails I received when restarting NiFi or killing the NiFi process:

  • Start event

Title: NiFi Started on Host pvillard-hdf-2 (


Apache NiFi has been started on host pvillard-hdf-2 ( at 2017/04/27 22:15:01.376 by user nifi

  • Stop event

Title: NiFi Stopped on Host pvillard-hdf-2 (


Apache NiFi has been told to initiate a shutdown on host pvillard-hdf-2 ( at 2017/04/27 22:14:35.702 by user nifi

  • Dead event

Title: NiFi Died on Host pvillard-hdf-2 (


It appears that Apache NiFi has died on host pvillard-hdf-2 ( at 2017/04/28 09:52:01.973; automatically restarting NiFi

OK, now let’s see how to configure the HTTP notification service. Let’s say I have a web server listening here:


My XML configuration is now:

        <property name="SMTP Hostname"></property>
        <property name="SMTP Port">587</property>
        <property name="SMTP Username"></property>
        <property name="SMTP Password">myPassword</property>
        <property name="SMTP TLS">true</property>
        <property name="From">"NiFi Service Notifier"</property>
        <property name="To">email address that will receive the notifications</property>
        <property name="URL">http://pvillard-hdf-4:9999/notification</property>

And my bootstrap.conf configuration file now contains:

# Notification Services for notifying interested parties when NiFi is stopped, started, dies

# XML File that contains the definitions of the notification services

# In the case that we are unable to send a notification for an event, how many times should we retry?

# Comma-separated list of identifiers that are present in the; which services should be used to notify when NiFi is started?,http-notification

# Comma-separated list of identifiers that are present in the; which services should be used to notify when NiFi is stopped?,http-notification

# Comma-separated list of identifiers that are present in the; which services should be used to notify when NiFi dies?,http-notification

I now restart NiFi, and I can confirm that I receive a notification (I used a standalone NiFi to confirm the reception of the notification):

Screen Shot 2017-05-02 at 5.15.29 PM

The content of the notification is the same as with the email notification service. Note that it’s also possible to configure properties for a keystore and a truststore to send notifications using HTTPS:


   /* The URL to send the notification to */
   <property name="URL"></property>

   /* Max wait time for connection to remote service - default is 10s */
   <property name="Connection timeout"></property>

   /* Max wait time for remote service to read the request sent - default is 10s */
   <property name="Write timeout"></property>

   /* The fully-qualified filename of the Truststore */
   <property name="Truststore Filename"></property>

   /* The Type of the Truststore. Either JKS or PKCS12 */
   <property name="Truststore Type"></property>

   /* The password for the Truststore */
   <property name="Truststore Password"></property>

   /* The fully-qualified filename of the Keystore */
   <property name="Keystore Filename"></property>

   /* The Type of the Keystore. Either JKS or PKCS12 */
   <property name="Keystore Type"></property>

   /* The password for the Keystore */
   <property name="Keystore Password"></property>

   /* The password for the key. If this is not specified, but the Keystore Filename, Password, and Type are specified, then the Keystore Password will be assumed to be the same as the Key Password */
   <property name="Key Password"></property>

   /* The algorithm to use for this SSL context. Either TLS or SSL */
   <property name="SSL Protocol"></property>

Note that I also received an email notification since I specified the two notifiers in the configuration.

Again, if you need a custom notifier, this should not be really difficult and you are more than welcome to contribute your notifier code to the Apache community. It will give more options to users when setting a bootstrap notifier.

As usual feel free to ask questions and comment this post.

Monitoring NiFi – Introduction

Apache NiFi 1.2.0 has just been released with a lot of very cool new features… and I take this opportunity to start a series of articles around monitoring. This is a recurring subject and I often hear the same questions. This series won’t provide an exhaustive list of the ways you can use to monitor NiFi (with or without HDF) but, at least, it should get you started!

Here is a quick summary of the subjects that will be covered:

For this series of article, I will use, as a demo environment, a 4-nodes HDF cluster (running on CentOS 7):

I’m using HDF to take advantage of Ambari to ease the deployment but this is not mandatory for what I’m going to discuss in the articles (except for stuff around the Ambari reporting task obviously).

I will not cover how setting up this environment but if this is something you are looking after, feel free to ask questions (here or on the Hortonworks Community Connection) and to have a look into Hortonworks documentation about HDF.

I don’t want to write a single (very) long article and for the sake of clarity there is one article per listed subject. Also, I’ll try to update the articles to stick as best as possible to latest features provided by NiFi over time.

Also, if you feel that some subjects should be added to the list, let me know and I’ll do my best to cover other monitoring-related questions.

List/Fetch pattern and Remote Process Group in Apache NiFi

I do see a lot of questions about how is working the List[X]/Fetch[X] processors and how to load balance the data over the nodes of a NiFi cluster once the data is already in the cluster. Since the question comes up quite often, let’s discuss the subject and let’s try to understand how things are working here.

I will assume that you are running a NiFi cluster since there is no problem about data balancing with a standalone instance 😉

The first thing to understand is: when running a cluster, one of the node is randomly designated as the “Primary node”. The election takes place when the cluster starts, and there is no way to decide which node will be the primary node. OK… you could force things when your cluster starts but there is no point to do such a thing if you want real high availability. So short line is: all nodes may have to be the Primary node at one point and don’t assume that the Primary node will be a given node in particular.

Second thing to understand is: the flow that you are designing on your canvas is running on all the nodes independently. Each node of the cluster is responsible of its own data and a relationship between two processors does not mean that the data going into this relationship will be balanced over the nodes. Unless you use a Remote Process Group (see below) the data will remain on the same node from the beginning to the end of the flow.

I will use a the following example to illustrate my explanations: I want to get files from a remote SFTP server and put the files into HDFS.

  • First idea (bad idea!) / GetSFTP -> PutHDFS

Screen Shot 2017-02-23 at 11.04.42 AM.png

The first option could be the pattern Get/Put which is perfectly fine with a standalone instance. However this will cause issues if you have a NiFi cluster. Remember? The flow is running on all hosts of your cluster. Problem is that you will have concurrent accesses from your nodes to the same files on the SFTP server… and if the processor is configured to delete the file once retrieved (default behavior) you will have errors showing up. Conclusion: always have in mind that a processor is running on all the nodes and can cause concurrent access errors depending on the remote system.

  • Second idea (not efficient!) / GetSFTP on Primary Node -> PutHDFS

The second option is to configure the GetSFTP processor to only run on the Primary Node (in the Scheduling tab of the processor configuration):

Screen Shot 2017-02-23 at 11.10.12 AM.png

This way, you will solve the concurrent accesses since only one node of your cluster (the Primary node) will run the GetSFTP processor.

Brief aside: remember, the flow is running on all the nodes, however if the processor is configured to run on the primary node only, the processor won’t be scheduled on nodes not being the primary node. That’s all.

With this approach the problem is that it’s not efficient at all. First reason is that you get data from only one node (this does not scale at all), and, in the end, only the primary node of your cluster is actually handling the data. Why? Because, unless you explicitly use a remote process group, the data will remain on the same node from the beginning to the end. In this case, only the primary node will actually get data from SFTP server and push it into HDFS.

  • Recommended pattern : ListSFTP -> RPG / Input Port -> FetchSFTP -> PutHDFS

To solve the issues, the List/Fetch pattern has been developed and widely used for a lot of processors. The idea is the following: the List processor will only list the data to retrieve on the remote system and get the associated metadata (it will not get the data itself). For each listed item, a flow file (with no content) will be generated and attributes will be populated with the metadata. Then the flow file is sent to the Fetch processor to actually retrieved the data from the remote system based on the metadata (it can be the path of the file on the remote system for example). Since each flow file contains the metadata of a specific item on the remote system, you won’t have concurrent accesses even if you have multiple Fetch processors running in parallel.

Obviously the List processor is meant to be run on the Primary node only. Then you have to balance the generated flow files over the nodes so that the Fetch processor on each node is dealing with flow files. For this purpose you have to use a Remote Process Group.

A Remote Process Group is an abstract object used to connect two NiFi setup together (the communication between the two NiFi is what we call Site-to-Site or S2S). It can be a MiNiFi instance to a NiFi cluster, a NiFi cluster to another NiFi cluster, a NiFi standalone to a NiFi cluster, etc. And it can also be used to connect a NiFi cluster to itself! This way the flow files will be balanced over all the nodes of the cluster. Few things to know with a Remote Process Group:

  1. You need to have an input port on the remote instance you are connecting to (in our case, you need a remote input port on your canvas).
  2. The address you give when configuring your remote process group does not matter in terms of high availability: once the connection is established with one of the nodes of the remote instance, the remote process group will be aware of all the nodes of the remote instance and will manage the case where the node specified in the address goes down.
  3. Your instances need to be configured to allow remote access. The required properties are:

# Site to Site properties

In the case of our SFTP example, it looks like:

Screen Shot 2017-02-23 at 11.47.40 AM.png

Let’s try to understand what is going on from a cluster perspective. Here is what we have in the case of a 3-nodes NiFi cluster with ListSFTP running on the primary node only:

Screen Shot 2017-02-23 at 12.03.22 PM.png

The ListSFTP when scheduled is going to list the three files on my remote SFTP server and will generate one flow file for each remote file. Each flow file won’t have any content but will have attributes with metadata of the remote files. In the case of ListSFTP, I’ll have (check the documentation at the “Write attributes” paragraph):

Name Description The hostname of the SFTP Server
sftp.remote.port The port that was connected to on the SFTP Server
sftp.listing.user The username of the user that performed the SFTP Listing
file.owner The numeric owner id of the source file The numeric group id of the source file
file.permissions The read/write/execute permissions of the source file
file.lastModifiedTime The timestamp of when the file in the filesystem waslast modified as ‘yyyy-MM-dd’T’HH:mm:ssZ’
filename The name of the file on the SFTP Server
path The fully qualified name of the directory on the SFTP Server from which the file was pulled

The ListSFTP processor will generate 3 flow files and, for now, all flow files are only on the primary node:

Screen Shot 2017-02-23 at 1.59.43 PM.png

Now the Remote Process Group has been configured to connect to the cluster itself, and I set the relationship going from ListSFTP to the Remote Process Group to connect with the input port I created (you may have multiple input ports in the remote system to connect with and you can choose the input port to connect to, that’s up to your needs). When the RPG (Remote Process Group) has the communication enabled, the RPG is aware of the three nodes and will balance the data to each remote node (be aware that there is a lot of parameters for Site-to-Site to improve efficiency). In my case that would give something like:

Screen Shot 2017-02-23 at 1.59.55 PM.png

Note: that would be an ideal case in terms of balancing but, for efficiency purpose, the Site-to-Site mechanism might send batch of flow files to the remote node. In the above example, with only 3 flow files, I would probably not end up with one flow file per node.

Now, since we have everything in the attributes of our flow files, we need to use the Expression Language to set the properties of the FetchSFTP processor to use the attributes of the incoming flow files:


This way, each instance of the FetchSFTP processor will take care of its own file (to actually fetch the content of the remote data) and there won’t be any concurrent access:

Screen Shot 2017-02-23 at 2.00.12 PM.png

All your nodes are retrieving data and you really can scale up your cluster depending on your requirements. Note also that the PutHDFS won’t be an issue neither since each node will write its own file.

As I said previously a lot of processors are embracing this pattern (and this is recommended way to use such processors with a NiFi cluster), and I’d strongly encourage you to do the same when developing your custom processors.

As always questions/comments are welcome.

Apache NiFi 1.1.0 – Secured cluster setup

Apache NiFi 1.1.0 is now out, and I want to discuss a specific subject in a couple of posts: how to scale up and down a NiFi cluster without loosing data? Before going into this subject, I want to setup a 3-nodes secured cluster using the NiFi toolkit. It will be my starting point to scale up my cluster with an additional node, and then scale down my cluster.

There are already great posts describing how to setup a secured cluster using embedded ZK and taking advantage of the NiFi toolkit for the certificates, so I won’t go in too much details. For reference, here are some great articles I recommend:

OK… let’s move on. My initial setup is the following:

  • My OS X laptop
  • 2 CentOS 7 VM

All nodes have required prerequisites (network, java, etc), have the NiFi 1.1.0 binary archive uncompressed available, and have the NiFi toolkit 1.1.0 binary archive uncompressed available.

On my OS X laptop, I will use the NiFi TLS toolkit in server mode so that it acts as a Certificate Authority that can be used by clients to get Certificates.

Here is the description of how to use the TLS toolkit in server mode:

./bin/ server

usage: org.apache.nifi.toolkit.tls.TlsToolkitMain [-a <arg>] [-c <arg>] [--configJsonIn <arg>] [-d <arg>] [-D <arg>] [-f <arg>] [-F] [-g] [-h] [-k <arg>] [-p
       <arg>] [-s <arg>] [-T <arg>] [-t <arg>]

Acts as a Certificate Authority that can be used by clients to get Certificates

 -a,--keyAlgorithm <arg>                   Algorithm to use for generated keys. (default: RSA)
 -c,--certificateAuthorityHostname <arg>   Hostname of NiFi Certificate Authority (default: localhost)
    --configJsonIn <arg>                   The place to read configuration info from (defaults to the value of configJson), implies useConfigJson if set.
                                           (default: configJson value)
 -d,--days <arg>                           Number of days issued certificate should be valid for. (default: 1095)
 -D,--dn <arg>                             The dn to use for the CA certificate (default: CN=YOUR_CA_HOSTNAME,OU=NIFI)
 -f,--configJson <arg>                     The place to write configuration info (default: config.json)
 -F,--useConfigJson                        Flag specifying that all configuration is read from configJson to facilitate automated use (otherwise configJson will
                                           only be written to.
 -g,--differentKeyAndKeystorePasswords     Use different generated password for the key and the keyStore.
 -h,--help                                 Print help and exit.
 -k,--keySize <arg>                        Number of bits for generated keys. (default: 2048)
 -p,--PORT <arg>                           The port for the Certificate Authority to listen on (default: 8443)
 -s,--signingAlgorithm <arg>               Algorithm to use for signing certificates. (default: SHA256WITHRSA)
 -T,--keyStoreType <arg>                   The type of keyStores to generate. (default: jks)
 -t,--token <arg>                          The token to use to prevent MITM (required and must be same as one used by clients)

In my case, I run the TLS toolkit on my node-3, and I run the following command:

./bin/ server -c node-3 -t myTokenToUseToPreventMITM -p 9999

On each of my node, I created a nifi directory in /etc and I ran a command using the toolkit to get my signed certificates generated into my current directory:

.../bin/ client -c node-3 -t myTokenToUseToPreventMITM -p 9999

And I have the following generated files on each of my nodes:

  • config.json
  • keystore.jks
  • nifi-cert.pem
  • truststore.jks

I now configure ZooKeeper as described here. Here is a short list of the tasks (in bold what slightly changed in comparison with my previous post):

  • configure conf/ to list the ZK nodes
  • configure ZK state ID file
  • set
  • set nifi.zookeeper.connect.string
  • set
  • set
  • set nifi.cluster.node.address=node-<1-3>
  • set nifi.cluster.node.protocol.port=9998
  • set<1-3>
  • set
  • set nifi.remote.input.socket.port=9997
  • set<1-3>
  • set nifi.web.https.port=8443

(ports are changed to ensure there is no conflict with the TLS toolkit running the CA server)

I now configure the following properties with what has been generated by the toolkit:

For one of my node, it gives:

Then I generate a certificate for myself as a client to be able to authenticate against NiFi UI:

.../bin/ client -c node-3 -t myTokenToUseToPreventMITM -p 9999 -D "CN=pvillard,OU=NIFI" -T PKCS12

Don’t forget the -T option to get your client certificate in a format that is easy to import in your browser (PKCS12). This command also generates a nifi-cert.pem file that corresponds to the CA certificate, you will need to import it in your browser as well (and you might need to manually update the trust level on this certificate to ensure you have access to the UI).

At this point I’m able to fill the authorizers.xml file. I need to specify myself as initial admin identity (to access the UI with full administration rights), and specify each nodes of my cluster (using the DN provided with the generated certificates). It gives:

        <property name="Authorizations File">./conf/authorizations.xml</property>
        <property name="Users File">./conf/users.xml</property>
        <property name="Initial Admin Identity">CN=pvillard, OU=NIFI</property>
        <property name="Legacy Authorized Users File"></property>

        <property name="Node Identity 1">CN=node-1, OU=NIFI</property>
        <property name="Node Identity 2">CN=node-2, OU=NIFI</property>
        <property name="Node Identity 3">CN=node-3, OU=NIFI</property>

WARNING – Please be careful when updating this file because identities are case-sensitive and blank-sensitive. For example, even though I specified

-D "CN=pvillard,OU=NIFI"

when executing the command to generate the certificates, it introduced a white space after the comma. The correct string to use in the configuration file is given in the output of the TLS toolkit when executing the command.

Once I’ve updated this file on each node, I’m now ready to start each node of the cluster.

./bin/ start && tail -f ./logs/nifi-app.log

Once all nodes are correctly started, I am now able to access the NiFi UI using any of the nodes in the cluster:

NiFi UI / 3-nodes secured cluster
NiFi UI / 3-nodes secured cluster

That’s all! Next post will use the current environment as a starting point to demonstrate how to scale up/down a NiFi cluster.

Apache NiFi 1.0.0 – Cluster setup

As you may know a version 1.0.0-BETA of Apache NiFi has been released few days ago. The upcoming 1.0.0 release will be a great moment for the community as it it will mark a lot of work over the last few months with many new features being added.

The objective of the Beta release is to give people a chance to try this new version and to give a feedback before the official major release which will come shortly. If you want to preview this new version with a completely new look, you can download the binaries here, unzip it, and run it (‘./bin/ start‘ or ‘./bin/run-nifi.bat‘ for Windows), then you just have to access http://localhost:8080/nifi/.

The objective of this post is to briefly explain how to setup an unsecured NiFi cluster with this new version (a post for setting up a secured cluster will come shortly with explanations on how to use a new tool that will be shipped with NiFi to ease the installation of a secured cluster).

One really important change with this new version is the new paradigm around cluster installation. From the NiFi documentation, we can read:

Starting with the NiFi 1.0 release, NiFi employs a Zero-Master Clustering paradigm. Each of the nodes in a NiFi cluster performs the same tasks on the data but each operates on a different set of data. Apache ZooKeeper elects one of the nodes as the Cluster Coordinator, and failover is handled automatically by ZooKeeper. All cluster nodes report heartbeat and status information to the Cluster Coordinator. The Cluster Coordinator is responsible for disconnecting and connecting nodes. As a DataFlow manager, you can interact with the NiFi cluster through the UI of any node in the cluster. Any change you make is replicated to all nodes in the cluster, allowing for multiple entry points to the cluster.


OK, let’s start with the installation. As you may know it is greatly recommended to use an odd number of ZooKeeper instances with at least 3 nodes (to maintain a majority also called quorum). NiFi comes with an embedded instance of ZooKeeper, but you are free to use an existing cluster of ZooKeeper instances if you want. In this article, we will use the embedded ZooKeeper option.

I will use my computer as the first instance. I also launched two virtual machines (with a minimal Centos 7). All my 3 instances are able to communicate to each other on requested ports. On each machine, I configure my /etc/hosts file with: node-3 node-2 node-1

I deploy the binaries file on my three instances and unzip it. I now have a NiFi directory on each one of my nodes.

The first thing is to configure the list of the ZK (ZooKeeper) instances in the configuration file ‘./conf/‘. Since our three NiFi instances will run the embedded ZK instance, I just have to complete the file with the following properties:


Then, everything happens in the ‘./conf/‘. First, I specify that NiFi must run an embedded ZK instance, with the following property:

I also specify the ZK connect string:


As you can notice, the ./conf/ file has a property named dataDir. By default, this value is set to ./state/zookeeper. If more than one NiFi node is running an embedded ZK, it is important to tell the server which one it is.

To do that, you need to create a file name myid and placing it in ZK’s data directory. The content of this file should be the index of the server as previously specify by the server. property.

On node-1, I’ll do:

mkdir ./state
mkdir ./state/zookeeper
echo 1 > ./state/zookeeper/myid

The same operation needs to be done on each node (don’t forget to change the ID).

If you don’t do this, you may see the following kind of exceptions in the logs:

Caused by: java.lang.IllegalArgumentException: ./state/zookeeper/myid file is missing

Then we go to clustering properties. For this article, we are setting up an unsecured cluster, so we must keep:

Then, we have the following properties:
nifi.cluster.node.connection.timeout=5 sec sec

I set the FQDN of the node I am configuring, and I choose the arbitrary 9999 port for the communication with the elected cluster coordinator. I apply the same configuration on my other nodes:
nifi.cluster.node.connection.timeout=5 sec sec

nifi.cluster.node.connection.timeout=5 sec sec

We have configured the exchanges between the nodes and the cluster coordinator, now let’s move to the exchanges between the nodes (to balance the data of the flows). We have the following properties:
nifi.remote.input.http.transaction.ttl=30 sec

Again, I set the FQDN of the node I am configuring and I choose the arbitrary 9998 port for the Site-to-Site (S2S) exchanges between the nodes of my cluster. The same applies for all the nodes (just change the host property with the correct FQDN).

It is also important to set the FQDN for the web server property, otherwise we may get strange behaviors with all nodes identified as ‘localhost’ in the UI. Consequently, for each node, set the following property with the correct FQDN:

And that’s all! Easy, isn’t it?

OK, let’s start our nodes and let’s tail the logs to see what’s going on there!

./bin/ start && tail -f ./logs/nifi-app.log

If you look at the logs, you should see that one of the node gets elected as the cluster coordinator and then you should see heartbeats created by the three nodes and sent to the cluster coordinator every 5 seconds.

You can connect to the UI using the node you want (you can have multiple users connected to different nodes, modifications will be applied on each node). Let’s go to:


Here is what it looks like:

Screen Shot 2016-08-13 at 7.33.08 PM

As you can see in the top-left corner, there are 3 nodes in our cluster. Besides, if we go in the menu (button in the top-right corner) and select the cluster page, we have details on our three nodes:

Screen Shot 2016-08-13 at 7.35.28 PM

We see that my node-2 has been elected as cluster coordinator, and that my node-3 is my primary node. This distinction is important because some processors must run on a unique node (for data consistency) and in this case we will want it to run “On primary node” (example below).

We can display details on a specific node (“information” icon on the left):

Screen Shot 2016-08-13 at 7.35.48 PM

OK, let’s add a processor like GetTwitter. Since the flow will run on all nodes (with balanced data between the nodes), this processor must run on a unique processor if we don’t want to duplicate data. Then, in the scheduling strategy, we will choose the strategy “On primary node”. This way, we don’t duplicate data, and if the primary node changes (because my node dies or gets disconnected), we won’t loose data, the workflow will still be executed.

Screen Shot 2016-08-13 at 7.45.19 PM

Then I can connect my processor to a PutFile processor to save the tweets in JSON by setting a local directory (/tmp/twitter):

Screen Shot 2016-08-13 at 7.52.25 PM

If I run this flow, all my JSON tweets will be stored on the primary node, the data won’t be balanced. To balance the data, I need to use a RPG (Remote Process Group), the RPG will connect to the coordinator to evaluate the load of each node and balance the data over the nodes. It gives us the following flow:

Screen Shot 2016-08-13 at 8.00.26 PM

I have added an input port called “RPG”, then I have added a Remote Process Group that I connected to ” http://node-2:8080/nifi ” and I enabled transmission so that the Remote Process Group was aware of the existing input ports on my cluster. Then in the Remote Process Group configuration, I enabled the RPG input port. I then connected my GetTwitter to the Remote Process Group and selected the RPG input port. Finally, I connected my RPG input port to my PutFile processor.

When running the flow, I now have balanced data all over my nodes (I can check in the local directory ‘/tmp/twitter‘ on each node).

That’s all for this post. I hope you enjoyed it and that it will be helpful for you if setting up a NiFi cluster. All comments/remarks are very welcomed and I kindly encourage you to download Apache NiFi, to try it and to give a feedback to the community if you have any.